parasj / contracode Goto Github PK

I tried to learn both pretrain and fine-tuning with one RTX-2080ti, but it takes a lot of time. What kind of learning environment did you use?
I would appreciate it if you could tell me the specs and number of gpu's you used for model pretraining and fine-tuning.

great work! I need some help of your codeclone dataset. If you do not mind spend a little time and help me figure out it , I will be very appreciated to you. I download it by the "scripts/download_data.py" in your repo (codeclone/full_data.json.gz) , but I do not know wether it is the dataset used in "4.1 Evaluating Functionality and Robustness: Zero-shot Code Clone Detection" in your paper. I see the "split" function in "representjs/clone_detection.py", so I'm confused... And for the 2065 pairs you mention in your paper, ( also in 4.1) , is it from the same dataset? and how to get it? If you do not mind spend a little time and help me figure out it , I will be very appreciated to you.

@parasj Is it your code applicable to Python language function? More precisely can the automated source to source compiler transformation be used for Python beyond js language?

Hi Parasj, thanks so much for your great work and released code, I notice that in the pre-trained step, the model is trained on the javascript_augmented.pickle.gz file. I want to ask that could we generate the augmented our js file by ourselves? If so, how to generate them? Thanks for your response and gudiance, best regards.

First of all, thanks for sharing the impressive work!

We were trying to get the finetuning for downstream task working, but got the following issue

torch.nn.modules.module.ModuleAttributeError: 'DataParallel' object has no attribute 'encoder'

It's quite likely that we were using the incorrect version of Pytorch (or maybe other dependencies).

Would you kindly share the proper version of the dependency required to run the project?

Cheers

Failed to decompress the data.zip file in the cloud disk. Is there any solution

Hi,

Thanks for the wonderful work. May I ask a question, when I pretrain LSTM model with the default settings, the memory is overflow. My server has 180G RAM, so may I ask how much RAM needed for pretraining?

Thanks and best regards.

How many GPUs are used by this project?

I follow the README to

Download the data subfolder from [this Google Drive link](https://drive.google.com/drive/folders/153pZfKPcr1-l8VaDPys29b1ElGLuoq3M?usp=sharing) and place at the root of the repository. This folder contains training and evaluation data, vocabularies and model checkpoints.

However, I do not fine the checkpoints and only the data.zip.

How can I obtain the checkpoint?

Thanks

Hi Parasj, thanks for publishing your code. I want to ask what are the parameters of the equipment used in your experiment? I found that 16G memory is not enough if using the javascript_augmented.pickle.gz file.

Originally posted by @QZH-eng in #6 (comment)